RE: Finding packages to enable by default

[Drew Adams]

> I suspect we'll see most of these packages among the top ones
> again if more data is collected.

Sounds like you think that the data gathered this way is in fact
a representative sample, at least wrt most of the features reported.

You might be right, but why do you think so?

So far, I see no reason.  Whether it is these features in reports
today or different ones in reports tomorrow, why suppose that
(most) bug-report features are representative of user configs?

The purpose of gathering that feature data for bugs is, hopefully
(but alas probably not so commonly), to learn something about the
context that manifested the bug.

That context is not necessarily representative of the user's
typical config.  In fact, it might well be pared down to provide
the context of a minimal repro recipe being reported.  In short,
a bug report is not a feature report.

But again,

>> Not saying that the info you gather this way has no meaning.
>> Just saying that not too much should be read into it (e.g.,
>> suppositions that it is representative in any way).

and on that I guess we agree.